How Mapillary Imagery Is Used to Update Maps

Mapillary is used to update maps by combining street-level photos with computer vision: contributors capture images with any camera, the platform extracts map features from those images automatically, and the resulting data is edited into maps through tools like OpenStreetMap, HERE Map Creator, and ArcGIS. It suits mappers, GIS teams, and organizations that need current detail on roads and roadside objects—especially where aerial imagery is outdated or too coarse.

The core mechanism: street-level photos fill what aerial imagery misses

Aerial and satellite imagery shows rooftops and road outlines but not what sits at ground level: traffic signs, lane markings, curb details, shopfronts, and small road changes. Mapillary's stated approach is to let anyone capture street-level images and publish them into a collaborative global street view, so those ground-level details become visible and mappable.

The platform reports more than 3 billion street-level images. Two properties matter for map updating:

• Coverage where official sources are thin. Contributors can deliberately photograph areas that are missing from commercial street-view services.
• Recency. Because anyone can upload, a road that changed last month can appear in imagery without waiting for a scheduled capture drive.

How features get from images into maps

Mapillary describes a pipeline in which computer vision extracts map features from the imagery, reducing manual tracing. In practice, the workflow looks like this:

1. Capture — photograph streets with a phone, action camera, or dedicated mapping camera. Mapillary publishes camera recommendations, and the platform accepts multiple camera types.
2. Upload — images go up through the mobile or desktop apps, or the command-line upload tool for larger batches.
3. Automatic extraction — computer vision identifies objects such as traffic signs and other map features in the images.
4. Edit into a map — use the extracted features and imagery as reference while editing in a compatible editor.

The expected result at each step is visible: uploaded sequences appear on the map, detected features become candidate data, and edits based on them show up in the target map.

A concrete example

Suppose a new roundabout opens and the aerial basemap still shows the old junction. A contributor drives through it with a phone mounted on the dashboard and uploads the sequence. The imagery shows the new lane geometry and the give-way signs. A mapper opens the area in an editor, uses the street-level view to trace the new road shape, and tags the signs—details the aerial image could not supply.

Where the data goes: integrations

Mapillary's imagery and extracted data are designed to feed existing mapping tools rather than replace them. The site lists integrations with:

There are also developer-facing resources—Mapillary JS for embedding imagery, image datasets, and OpenSfM for structure-from-motion processing—for teams building their own tools on top of the data.

Who benefits, and under what conditions

• Individual mappers get free access to reference imagery for editing; the value depends on whether coverage exists in their area, which they can check by exploring the map before committing time.
• GIS and city teams benefit when they need current roadside detail for planning or navigation data; the open-data model is positioned to support decisions on roads, mobility, and autonomous driving.
• Organizations with blank spots can fill them by capturing imagery themselves rather than waiting for a provider.

Two conditions to check before relying on it: confirm actual image coverage for your target area on the map, and confirm that your editing tool of choice is among the supported integrations. The site does not publish pricing details in the material reviewed here, so treat cost and account requirements as something to verify directly rather than assuming free or login-free access.

Getting started

To try the workflow end to end: explore the map to find a gap or a recently changed area, download the app, capture and upload a short sequence, then open the same location in OpenStreetMap or another integrated editor and compare what the street-level view reveals against the existing map. That single loop demonstrates both halves of the system—collaborative capture and computer-vision-assisted editing—better than any description.